| Registration of drug trials |
| Zarin et al. 2005 (e6) |
Description of registration of studies before and after implementation of ICMJE regulations by inspection of register entries at ClinicalTrials.gov on 20 May 2005 and again on 11 October 2005 |
Considerable increase in registrations of clinical studies (from 13 153 to 22 714, also non-drug studies). Increase in registrations of interventional new drug studies from 2010 to 3516. Non-specific entries still found under “Name of intervention” on 11.10.2005, frequently (376/1247; 30%) no entry under “Primary outcome measure”. |
| Lott et al. 2006 (e7) |
Investigation of behavior of 31 pharmaceutical companies that had placed advertisements in two major dermatology journals with regard to registration of their clinical trials after implementation of ICMJE regulations and guidelines of pharmaceutical ‧associations, by means of questionnaires, monitoring of company websites, and inspection of registers at ClinicalTrials.gov and isrctn.org in year 2005/2006 |
Only 5 of 31 companies responded and one refused permission for publication of its answers. Proportion of trials registered varied from 0% to 100% at the remaining four companies. One company’s response: decision on registration made individually for each project, no formal guideline in force. Twenty of 31 companies had registered at least one study at ClinicalTrials.gov, 16 of 31 at isrctn.org. Nine pharmaceutical companies had entered no trial in either register. Twenty-three of 31 companies (74%) gave no information about registration of their clinical trials on their websites. |
| Publication of drug trials |
| Conen et al. 2008 (e25) |
Influence of study funding source on subsequent distribution of data in 303 consecutively published studies on cardiovascular ‧research topics, published between 2000 and 2005 |
Results of studies funded by pharmaceutical companies are cited more often. Among drug trials (n = 194), median number of citations was 43 for studies funded by pharmaceutical companies, 42 in the case of mixed funding, and 33 for studies financed by sources other than pharmaceutical companies (differences not statistically significant). Analysis of all studies shows a particularly clear difference for advantage of new treatment over existing standard, otherwise opposite finding. |
| Hole et al. 2009 (e23) |
Determination of number of completed investigations and of ‧published results among the 245 studies registered in archive of Norwegian National Committee for Medical Research Ethics in the year 2000 |
Investigation completed: 178/245 (73%); results published in journal: 131/245 (54%); published as abstract or as report to sponsor: 34/245 (14%); not published at all: 80/245 (33%). Pharmaceutical companies as sponsors seem to be conducive to both study completion and publication of results in specialist journals. |
| Jefferson et al. 2009 (e26) |
Analysis of agreement between study results and conclusions (concordance), funding, and distribution of 259 studies on the ‧effect of influenza vaccinations |
Studies completely or partly funded by pharmaceutical companies are more frequently published in high-ranking journals and cited more often than studies with state or private funding; this cannot be explained by quality or size of study. |
| Kuriya et al. 2008 (e27) |
Quality of press releases on results of original research by the 10 leading pharmaceutical companies worldwide; published on their website in 2005 |
Key trial data are reported in press releases, but limitations are rarely mentioned (6%). In 29% of cases no quantification of study results. In 47% an author is quoted, often stressing benefit. |
| Kho et al. 2009 (e22) |
Determination of rate of full publication of 109 abstracts on 86 trials of rituximab in non-Hodgkin’s lymphoma, presented at annual congress of a hematology society between 1997 and 2001; also determination of time to publication and of predictors for these parameters |
Publication rate of abstracts was 52.3%, median time to full publication 1.4 years. Affiliation of authors to pharmaceutical companies (OR 4.60, 95% CI 1.32–16.08) and type of presentation were independent predictors of full publication. Predictors for time to full publication were not identified. |
| Krzyzanowska et al. 2003 (e21) |
Determination of factors that influence time to publication in 510 abstracts of large RCTs (≥200 patients) in oncology; presented at annual meeting of a society between 1989 and 1998 |
Studies funded by pharmaceutical companies (74/510; 15%) were published more rapidly than those financed by study group (294/510; 58%, including 17 supported by pharmaceutical companies) or those with no information on funding (142/510; 28%). After 5 years 17% of the studies funded by pharmaceutical companies, 27% of those financed by the study group, and 29% of those with no reported sponsor remained unpublished (p = 0.02). |
| Lee et al. 2008 (e14) |
Determination of publication status and factors associated with publication in trials of drugs licensed by FDA between 1998 and 2000 |
Of 909 licensing studies for 90 newly licensed drugs (89 from pharmaceutical companies), 43% (394/909) had been published after 5 years, including 76% (257/340) of pivotal trials. Likelihood of being published was twice as high for studies with statistically significant results than for those with non-significant results. Size of sample was also positively associated with likelihood of publication. |
| Melander et al. 2003 (e17) |
Importance of selective and multiple publication in 42 RCTs on five SSRIs that had been submitted to the Swedish licensing authority for commercial licensing between 1983 and 1999 |
Multiple publication: 21 trials resulted in at least two publications, three trials to five publications each. Selective publication: independent publication more frequent in trials with significant results. Many studies reported more favorable per-protocol analysis, rather than intention-to-treat analysis. |
| Rising et al. 2008 (e15) |
Determination of publication rate of efficacy studies presented to FDA in 2001–2002 (n = 164) and comparison with corresponding publications |
By 2007, 78% (128/164) of studies had been published, with a significantly higher likelihood for those with favorable results and active controls. Publications featured more positive results than FDA documents, e.g., by addition or deletion of findings or alteration of statistical significance or conclusions. |
| Ross et al. 2009 (e24) |
To check completeness of registration and extent of selective publication, data from register at ClinicalTrials.gov were compared with corresponding publications (registration after 31 December 1999, study ended by 08 June 2007) |
Overall, fewer than half of studies investigated were published (311/677; 46%). Clinical trials funded by pharmaceutical companies (144/357; 40%) were less likely to be published than those funded neither by pharmaceutical companies nor by state support (110/198; 56%; p<0.001), but there was no difference from state-funded trials (57/122; 47%; p = 0,22). |
| Turner et al. 2008 (e16) |
Selective publication in 74 studies on 12 antidepressives licensed between 1987 and 2004; submitted to American ‧licensing authority |
Thirty-seven of 38 studies with positive result published, 3/36 studies with negative result adequately published, 11 studies published with negative result presented as positive. Separate meta-analyses of data provided to FDA and published data showed greater strength of effect in published data (range 11% to 69%, overall mean 32%). |
| Vedula et al. 2009 (e18) |
Analysis of presentation of study results by comparison of pharmaceutical manufacturers’ internal documents with corresponding publications of manufacturer-funded trials of gabapentin for non-licensed (“off-label”) indications |
In 8 of 12 published studies, primary endpoint in publication differed from that described in protocol. Studies with non-significant results (p≥0.05) for primary endpoint defined in protocol were published either incompletely or with altered primary endpoint. Of 21 endpoints described in protocols as primary, four were published as secondary endpoints and six not at all. Of 28 primary endpoints published, 12 were newly introduced. |
| Von Elm et al. 2008 (e20) |
Factors associated with publication in 451 study protocols and 233 corresponding articles on RCTs; submitted to Swiss national ethics committee between 1988 and 1998 |
Funding independent of pharmaceutical companies is associated with publication (OR 2.42, 95% CI 1.14–5.17), as are multicenter status, international cooperation, and case number >236. |
| Whittington et al. 2004 (e19) |
Comparison of published (five RCTs) and unpublished data on SSRIs in children and adolescents; survey of literature up to 2003 |
Published data suggest a positive benefit/harm balance for some SSRIs (paroxetine, sertraline), whereas risk predominates when unpublished data are taken into account (exception: fluoxetine). |
| Publication rights and control of trial data |
| Gotzsche et al. 2006 (e28) |
Investigation of publication rights of all 44 RCTs initiated and published by pharmaceutical companies in a region of Denmark in 1994 and 1995 |
Sponsors had right of access to data in 16/44 protocols, and 16 empowered them to end the study for any reason at any time. Limitations on publication rights in 40/44 protocols; sponsors granted rights over data, manuscript release, or both in 22. With one exception, sponsors’ rights not noted in publication. |
| Henry et al. 2005 (e29) |
Characterization of relationship between pharmaceutical companies and medical specialists in Australia with regard to accomplishment of studies by means of questionnaires in 2002 and 2003 |
Response rate 823/2120 (39%). Most frequent negative experience: premature discontinuation of study (114/823; 14%), in some cases on commercial grounds; also drafting of manuscript by pharmaceutical company staff (100/823; 12%). Study results delayed, distorted, or not published at all. Some 21% of physi‧cians participating in research projects report grave deficiencies. |
| Authorship—ghostwriters and guest authors |
| Gotzsche et al. 2007 (e4) |
Investigation of prevalence and nature of ghostwriting by comparing protocol and publication of all studies initiated by pharmaceutical companies in a region of Denmark in 1994 and 1995 |
Signs of involvement of ghostwriters in 33 of a total of 44 studies (75%; 95% CI 60–87%). In 31/44 publications (70%) statistical analysis was carried out by pharmaceutical company staff who were not named as authors. |
| Ross et al. 2008 (e5) |
Description of type and extent of ghostwriting and guest authorship in the case of rofecoxib, by inspection of publications and pharmaceutical companies’ internal documents that were provided in the context of legal proceedings |
Manuscripts for publications of clinical studies and reviews were written by company staff or contractors, then paid guest authors were recruited and frequently credited as first or second author. Support by a pharmaceutical company was mentioned in 22 (92%) of 24 publications of clinical studies but in only 36 (50%) of 72 reviews. |
